chore(api): stop injecting SessionId into merged spec

[hotdata-automation]

Auto-generated from the updated HotData OpenAPI spec.
Source: https://github.com/hotdata-dev/www.hotdata.dev/pull/283

[claude]

Dropping the X-Session-Id block here (and in results_api::get_result / list_results) leaves the SDK internally inconsistent, because the hand-written mirrors of these same requests still send it:

• src/query.rs:618 (send_query, the path Client::query actually uses) → sends X-Session-Id
• src/client.rs:468 (Client::submit_query) → sends X-Session-Id
• src/arrow.rs:305 (apply_apikey_headers, used by get_result_arrow / stream_result_arrow) → sends X-Session-Id

After this change there is no X-Session-Id emission left anywhere in src/apis, so a single Client::query() call now submits with the session header (send_query) and then polls GET /v1/results/{id} without it (wait_for_result → generated get_result). Same for JSON vs Arrow result fetches: the Arrow path is session-scoped, the generated path is not.

This also silently narrows public API surface: ClientBuilder::session_id (src/client.rs:126) and the HOTDATA_SESSION_ID env var still install the key into configuration.api_keys and their docs still say "Installed as the X-Session-Id header when present" — which is now only true for the hand-written paths. If the backend still scopes results by session, a session-scoped client's get_result / list_results calls are no longer scoped.

Please resolve one way or the other before merge:

1. If session auth is genuinely gone from the API — finish the removal in the same PR: drop ClientBuilder::session_id, SESSION_ID_HEADER, the HOTDATA_SESSION_ID lookup, the SESSION_ID_HEADER entries in src/query.rs / src/arrow.rs / src/client.rs::submit_query, the now-wrong arrow.rs tests and doc comments that say "mirroring the generated get_result" (src/arrow.rs:301, src/arrow.rs:496, src/query.rs:639), and the X-Session-Id rationale in .openapi-generator-templates/README.md:7. Record it in the changelog as Breaking since session_id() is public API.
2. If session scoping is still enforced server-side — the spec change is wrong and this regen should not ship.

[claude]

nit: (not blocking) this entry is the raw commit subject and describes the spec pipeline ("merged spec"), not what changes for a user of this crate. Every other entry in this file is written from the caller's point of view. Something like "query, get_result, and list_results no longer send the X-Session-Id header" would tell a reader whether they are affected.

[claude]

Review

Blocking Issues

• src/apis/query_api.rs:57, src/apis/results_api.rs:76, src/apis/results_api.rs:150 — removing the X-Session-Id emission leaves no session-header emission anywhere in src/apis, while the hand-written mirrors of the same requests still send it: src/query.rs:618 (send_query), src/client.rs:468 (submit_query), src/arrow.rs:305 (Arrow result fetch). Net effect: one Client::query() call now submits with the session header and then polls GET /v1/results/{id} without it, and JSON vs Arrow result fetches disagree on scoping.
• src/client.rs:126 — ClientBuilder::session_id and the HOTDATA_SESSION_ID env var are public API. They still install the key and still document "Installed as the X-Session-Id header when present", which is now only true on the hand-written paths. If the backend still scopes results by session, session-scoped get_result / list_results calls are silently no longer scoped.
• Stale statements left behind: src/arrow.rs:301 and src/query.rs:639 say the helpers mirror the generated ops' isKeyInHeader blocks; the test at src/arrow.rs:496 asserts behavior described as "matching the generated get_result". Neither holds after this change. .openapi-generator-templates/README.md:7 still cites X-Session-Id as the reason for the template override.

Action Required

Pick one and land it in this PR:

1. Session auth is gone from the API → finish the removal: drop ClientBuilder::session_id, SESSION_ID_HEADER, the HOTDATA_SESSION_ID lookup, the session entries in src/query.rs / src/arrow.rs / Client::submit_query, and the now-wrong tests and doc comments listed above. Record it as a Breaking changelog entry, since session_id() is public.
2. Session scoping is still enforced server-side → the upstream spec change is wrong and this regen should not ship; fix it at the source (the linked www.hotdata.dev PR 283).

Also see the non-blocking changelog-wording comment.